Skip to content
10 min readCritique

Introducing Workspace: one place for review, chat, build, and repair

Critique is merging PR review, Chat, Builder, and Remedy into a single operating surface — one rail, one ledger, one always-on engineer.

One surface. Four jobs.

May 2026 · Workspace

critique.sh

Review

Was a separate surface

Chat

Was a separate surface

Builder

Was a separate surface

Remedy

Was a separate surface

Now one place

Workspace

One left rail. One credit ledger. One repo context that follows you from question to verdict to patch to build.

/workspace — chat history, builder runs, review context, remedy handoffs

Critique Workspace is Critique’s unified operating environment for software teams. It replaces the old pattern of opening Chat in one tab, Builder in another, and the dashboard for review runs in a third. You pick a repository once, work in one left rail, and move between explanation (Chat), enforcement (PR review), repair (Remedy), and execution (Builder) as modes — not separate products.

4→1
Surfaces merged — Review, Chat, Builder, Remedy
/workspace
Primary signed-in entry point (May 2026)
0
Credit ledger across chat, review, builder, and remedy runs
v0.3.0
Release that shipped unified shell, timeline activity, and usage truth

For the last year, Critique grew the way most serious products do — one sharp wedge at a time. Review landed on the pull request. Chat gave you repo-grounded answers between PRs. Remedy closed the gap between finding and fix. Builder exposed the same sandbox fabric for open-ended execution. Each surface worked. Together, they still felt like four rooms in the same house with four different door locks.

That fragmentation had a cost you probably felt even if you never named it. You explained the same bug twice — once in Chat, again when a review run started. You lost thread context when you jumped from a verdict to a Remedy attempt. Builder history lived in one rail while chat lived in another. The product rhymed architecturally; the interface did not.

The trigger was not aesthetic. It was behavioral. Teams using Critique heavily were already treating the product as a loop: ask in Chat, review on the PR, fix with Remedy, extend with Builder. The value compounded when context traveled with them. When it did not, the loop broke — and the break always happened at a tab boundary.

Before
Open ChatSwitch to DashboardFind review runJump to RemedyOpen BuilderRe-explain context
Workspace
Open /workspaceAsk or actReview context attachedRemedy from threadBuilder when readySame ledger throughout

Workspace removes the mode tax. Not by hiding Review, Chat, Builder, or Remedy — but by making them modes inside one session instead of destinations you navigate between. Same dark, quiet shell. Same left rail for recent chats and builder runs. Same composer when you need to say what should happen next.

Workspace

The signed-in working surface at /workspace. One rail for history, one context bundle for the active repo, and mode-aware actions that hand off between conversation, execution, and review without a hard reset.

/workspace
Not a replacement

PR review still posts to GitHub. Remedy still runs in isolated sandboxes. Builder still executes cloud jobs. Workspace is the control room — not a rewrite of the pipelines underneath.

GitHub checks unchanged

If you have used Critique Chat or Builder before, Workspace will feel familiar immediately. The difference is continuity. Open a thread, escalate to build mode with ?mode=build, return to the same rail, and pick up the review run that thread referenced. The surfaces were always related. Now the UI admits it.

Each mode still does a distinct job. Workspace does not blur them into a generic “AI assistant.” It keeps the boundaries sharp while sharing the infrastructure that was duplicated before: repo indexing, model routing, credit accounting, and session memory.

Hybrid pattern — where most teams land
Explain & explore
  • Repo-grounded Chat
  • Architecture questions
  • PR summaries
  • Memory across sessions
Shared in Workspace
  • One credit ledger
  • One history rail
  • One repo context
  • Handoffs without re-briefing
Execute & enforce
  • Multi-agent PR review
  • Remedy patch runs
  • Builder cloud jobs
  • GitHub check runs

Open /workspace, select a GitHub repository, and start in Chat or switch to build mode with ?mode=build. Recent chats and Builder runs appear in one left rail. When a PR review completes, open the run from the same surface. Start Remedy from the thread when a finding is bounded. Credits, usage, and model routing draw from the same ledger you see under Dashboard → Usage — there is no second billing story for Workspace.

Unified shell
  • /workspace as the shared operating surface for Chat and Builder history
  • One left rail for recent chats and builder runs
  • Builder styling aligned with the signed-in workspace instead of a separate console
  • Mode-aware actions — Build, Plan, and review handoffs from the same composer
Builder inside the loop
  • Searchable repository and model selectors
  • Activity rendered as a timeline, not raw log spam
  • Prompt-derived branch publish and draft PR reuse after a run
  • Microphone capture in the composer — dictate a build prompt from the workspace
Reliability underneath
  • Stalled OpenCode turns bounded with liveness checks
  • Per-call OpenRouter-style usage persisted for review and Builder paths
  • Credit reconciliation closer to source-of-truth accounting
  • Review run detail tuned for triage — verdict first, usage expandable
Morning
Ask before you open the diff

You land in /workspace, select the repo, and ask why the auth middleware behaves differently on staging. Chat pulls indexed context. No review credits burn for a question.

Midday
Review meets the thread

A teammate opens a PR. The GitHub App queues Critique review. You open the run from Workspace — the same rail that held your morning chat — and read the verdict alongside the evidence pack.

Afternoon
Remedy when the fix is bounded

Two findings are narrow: missing null guard, stale import. You start Remedy from the thread. The sandbox clones the head branch, patches, validates, and pushes — visible live without leaving Workspace.

Late
Builder for the work that is not a PR yet

You need a broader refactor — extract a shared module, update three callers. You switch to build mode, pick the model, and let the cloud job run. The activity timeline streams back. When it finishes, the branch and draft PR are ready.

That loop — explain, enforce, repair, extend — is the job Critique was always built for. Workspace is the first interface that treats it as one continuous shift instead of four product tabs.

Workspace unifies the signed-in experience — not every pipeline
SurfaceRoleWhere it lives
CheckpointPre-review trust gateDeterministic intake before credits burn/checkpoint · Dashboard → Checkpoint
PR ReviewMerge-gate verdict on GitHubScout, specialists, lead synthesisGitHub checks + Dashboard runs
WorkspaceSigned-in control roomChat, Builder, context, handoffs/workspace
DashboardOperator hubInstallations, policy, usage, help/dashboard

Checkpoint and automated review still run on the GitHub webhook path. Workspace is where humans stay oriented while those pipelines execute.

Start here if…
  1. 1
    You already bounce between Chat and Builder?
    Workspace is the default. Bookmark /workspace instead of juggling /chat and /builder.
  2. 2
    You want review context while you fix?
    Open the review run from the same rail, start Remedy, and keep the thread that explains why the patch exists.
  3. 3
    You only use the GitHub App today?
    Install stays the same. Add Workspace when you want repo Q&A and execution between PRs — the App and Workspace share installations and credits.
  4. 4
    You need enterprise policy controls first?
    Dashboard → Automation still owns review policy, Checkpoint rules, and model tiers. Workspace respects those settings; it does not replace them.
No. They are modes and pipelines inside Critique, not deleted products. /chat and /builder still resolve; /workspace is the unified entry point we recommend for signed-in work. Remedy still starts from review findings or an explicit Chat action when write access allows.
No for Chat and Builder — connect a repository in the dashboard and you can explore and execute. Yes for automated PR review, GitHub check runs, and Remedy pushes to PR branches. Workspace surfaces all of it when your installation is connected.
No separate SKU. Chat, review, Builder, and Remedy draw from the same credit pool. Lightweight Q&A remains cheaper than a full review pipeline; long Builder runs and Remedy attempts cost more. Dashboard → Usage shows the ledger.
It is still there. Workspace’s left rail lists recent chats and builder runs together. Nothing was deleted in the merge — the navigation model changed, not your data.
Those tools optimize for writing code on your machine. Critique optimizes for merge-gate quality and repo-scoped execution in the cloud. Workspace is where you orchestrate that — review evidence on the PR, fixes in sandboxes, questions grounded in your index — without pretending generation and review are the same job.
Deeper cross-mode context — review runs pinned to chat threads, richer memory handoffs, and tighter Builder ↔ Remedy bridges. The v0.3.0 release unified the shell and history. The next waves make the handoffs invisible.

Open Workspace

Sign in, pick a repo, and stay in one place from question to verdict to patch. Five hundred free credits still apply — no card required.

Go to Workspace